Research Assistant

• Outcome: Review
• Latest audit: May 29, 2026, 9:30 PM
• Version:
  1.0.0

Overview

This skill appears low-risk to your device, but it advertises real web research while generating fake sources and citations.

Install only if you understand this behaves like a prototype or report generator, not a real research tool. Do not rely on its citations, credibility labels, or fact-checking unless the implementation is replaced with real source retrieval and verification.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 95% confidence
• Finding: The skill advertises real web research, fact verification, and cited reporting, but the finding indicates it may instead fabricate search results, placeholder citations, and agent behavior. This is dangerous because users may rely on false or unverifiable outputs for decisions, while believing they were grounded in external sources and cross-checked evidence.

Intent-Code Divergence

Medium

• Confidence: 95% confidence
• Finding: The code explicitly simulates search results and never performs actual external retrieval, yet the skill is presented as a web research assistant. This creates a deceptive capability gap that can cause users or downstream agents to trust fabricated source discovery and citations as if they were grounded in real research.

Description-Behavior Mismatch

High

• Confidence: 99% confidence
• Finding: The skill claims web research and citation generation, but it fabricates sources using example.com URLs and synthesized summaries rather than retrieving real information. In a research context, this is especially dangerous because it can produce authoritative-looking but false citations, enabling misinformation, bad decisions, or fraudulent outputs that appear verified.

Missing User Warnings

Low

• Confidence: 82% confidence
• Finding: If the skill sends user topics or prompts to search engines or remote websites without warning, users may unknowingly disclose sensitive project names, personal data, or confidential research interests to third parties. In a research assistant context, this is plausible and increases privacy risk because user queries are likely to be forwarded externally during normal operation.

VirusTotal

65/65 vendors flagged this skill as clean.